I've been involved in several projects assessing wireless network performance over the last year, and I've gained an appreciation for the associated complexities. Testing has included both Wi-Fi hotspot networks and 3G networks. 3G network testing involved both CDMA 2000 EV-DO (Evolution Data Optimized), as offered by Sprint and Verizon, and HSDPA (High Speed Downlink Packet Access), as offered by Cingular. My most recent effort involved reviewing and advising on the testing done for the August 31 Network Computing cover story I wrote, "Taking Advantage of Wide-Area Wireless--It's Time to Decide"

One of the first factors you must consider is that in the past, the wireless network was usually the bottleneck. However, today's wireless networks are becoming so fast that other parts of the network may be the constraining factor, such as the backhaul between base station and operator core network, an item I wrote about last time when commenting on Sprint Nextel's WiMAX choice. Another bottleneck could be the speed of the Internet connection of the server you are accessing while mobile. Even if the speed of that connection is faster than the wireless connection, there may be other traffic over that connection limiting the amount of available bandwidth. Or there may be security gateways that can't keep up with the remote access network. These back-end connection issues actually forced us to redo quite a few of our tests for last month's Network Computing story, which initially showed HSDPA throughputs to be quite low (approximately 300 kbps) but then considerably higher (700 kbps to 1 Mbps) once we eliminated the server-side bottlenecks.

In a comparison of Wi-Fi hotspots vs. EV-DO--published in a free white paper available on my Web site--we took no chances with back-end limitations and tested against a dedicated server that we collocated at an Internap collocation site with a 100 Mbps connection. For throughput, I prefer doing FTP (File Transfer Protocol) tests using test files of 1 megabyte to 3 megabytes. It's also possible to use public Web sites that provide throughput measurements, but I don't consider these as dependable or as repeatable as testing against a known and controlled test server. Even with file transfer, however, there are complications. One of our tests was to load the network with four simultaneous downloads. With EV-DO we found it difficult to get the four downloads to occur at the same time; the latency of EV-DO caused problems when attempting to have client computers try to open the FTP data port at the same time. So we switched to using HTTP Get operations, which worked much better because of the simpler protocol involved and allowed us to demonstrate how just four simultaneous downloads on an EV-DO cell sector can affect throughput. The average throughput we measured for a single computer--430 kbps--went down to 248 kbps with four systems using the network.

File download tests measure one aspect of a wireless networks capability, but they don't capture aspects such as latency. One good test here is to load Web pages that are representative of Internet Web sites, for example, complex pages such as www.cnn.com. The challenge here is that public sites vary their content moment by moment and have varying usage loads. In our EV-DO vs. Wi-Fi hotspot testing, we solved this problem by building our own Web page on our test server with the same complexity as cnn.com, which has about a 100 graphical objects. With Web downloads, it's important to clear the caches before each download, otherwise only a portion of the page may be downloaded in subsequent tests. In our results, the hotspot network averaged 6 seconds for downloading the page versus EV-DO, which averaged 25 seconds, demonstrating how latency can affect some applications more than throughput. Another good test is to send and receive a certain number of e-mail messages, as this also is representative of real-world operation and captures the effects of both throughput and latency.

Another aspect of testing networks, especially cellular networks, is that voice and data loading can affect performance, and performance varies depending on signal strength and interference. Hence, it is not enough to measure in just one location. In fact, you need to measure in at least eight locations to obtain representative results. More locations would be even better, but costs really start running up. To complicate matters further, operator networks may use one vendor's radio-access network hardware in one market and another vendor's equipment in another market, with possibly different performance. I have seen significantly different throughput rates in different cities with some operator networks.

Measuring time is also a problem. For ad hoc testing you can use your watch, but for operations taking seconds, this method is not very accurate. In our testing, we built test scripts that would invoke all test operations and automatically capture the amount of time involved.

Measuring latency is not trivial either, as we learned in our recent HSDPA testing. In the past, our tests would use 11 successive pings, discarding the first one because that one can take extra time to bring up the data traffic channel on cellular networks. With HSDPA, however, we saw much higher latency results than we expected--300 msec to 400 msec. It was only after conferring with the operator that we learned that it takes a certain amount of traffic to invoke the high-speed data channel. Our ping packets traveled on a lower speed channel and were not representative of what an application would actually obtain. We needed to increase the ping test rate to twice a second. The DOS command-line ping command in Microsoft Windows only does pings once a second, but you can use a utility such as PingPlotter to increase the ping rate. Once we did this, latency results went way down, to as low as 130 msec when testing against google.com.

Bottom line: It's easy to do some quick and dirty tests to characterize a wireless network's performance. However, to comprehensively characterize these networks is a major undertaking.

top^

Sprint Nextel, WiMAX and the Mobile Broadband Conundrum

Sprint Nextel announced on August 8 its choice of WiMAX technology for its mobile broadband network that will operate in its 2.5 GHz Broadband Radio Service (BRS) spectrum. Every other wireless pundit is going to have something to say about this, and I have so much to say that the constraints of this column will be a challenge. First, I'm going to make some general comments about the company's technology choice and early claims. Then I'm going to zoom in on the crux of the challenge that Sprint Nextel will have to address: managing capacity and performance expectations while finding a way of charging for service that enables a successful business model. This will be a conundrum, which according to one definition I just read means "a paradoxical, insoluble or difficult problem."

To begin with, it's clear that Sprint Nextel's choice of Mobile WiMAX (based on the IEEE 802.16e-2005 standard) is a huge win for Intel--especially in conjunction with Clearwire, which also recently chose WiMAX. The technology has now achieved some credibility. Mobile WiMAX embodies many radio innovations and, on paper at least, has impressive capabilities. I do take exception, however, to labeling WiMAX a 4G technology. First, there is no official 4G standards work. But people have been researching 4G approaches for quite a few years, and commonly accepted requirements include 1 Gbps peak speed and 100 Mbps average speeds--50 times faster than what Sprint Nextel will be able to provide. However, there is some karmic balance in all of this, because Qualcomm, a WiMAX opponent, has been promoting CDMA2000 1xRTT for many years as 3G. 1xRTT has a peak network speed of 153 kbps. If 1xRTT is 3G, then that would definitely make Mobile WiMAX 4G. But the fact is, Mobile WiMAX will only yield a modest improvement over 3G systems of today, and the aggressive 3G roadmaps I described in my column three weeks ago will allow these systems to largely match Mobile WiMAX performance.

Now, let's get to the conundrum of capacity/performance/pricing. The first aspect is one of capacity. Regardless of how fast the radio technology is, most cell sites today in the United States are backhaul constrained by T1 circuits. A network architect commented to me last week, "Mobile WiMAX. That's just another fast airlink. When will people realize that the real issue is the wireline network?" Think about it. If a Mobile WiMAX system uses 10 MHz radio channels, and assuming 7.5 MHz is allocated to the downlink (this is a time division duplex system) and using a spectral efficiency value of 1 bps/Hz/sector (an optimistic value in itself for a loaded network with lots of interference), this is 7.5 Mbps of throughput per sector. With three sectors in a cell, this would require a whopping 15 T1 circuits. Clearly, this is not going to happen any time soon. Eventually, yes. 2008? No. This takes me to the next issue, which is performance. Sprint Nextel has indicated 2 Mbps to 4 Mbps average throughput. I view this as a very aggressive claim. First, the company will need all the backhaul capacity I described to support these rates. But even if Sprint Nextel had it, how many active users in a cell sector (one mile is typical cell site spacing) could the company support? Simply divide 7.5 Mbps by 2 Mbps to 4 Mbps, and you end up with a small number. Granted, not everybody is downloading at the same time, so the network can be oversubscribed, but only by so much before things slow down below the promised rates.

These are the same types of issues that cellular operators face. But what's different with 3G is that much of the data use is with bandwidth-efficient applications such as SMS and RIM BlackBerries. Business data users tend not to consume under 100 megabytes per month on average, so capacity has not been as much of an issue. But Sprint Nextel is emphasizing the consumer front, which is multimedia driven. I've seen dimensioning exercises for Mobile WiMAX with figures of 1 gigabyte per subscriber per month. Compared to 3G, this is lots of capacity--in fact, about 10 times the amount of a 1,000-minute voice user. But the problem is that 1 gigabyte is less than one DVD-quality movie download, even using MPEG4 compression.

And that's the conundrum. 3G networks can easily support voice and pictures, so phone service, e-mail and Web browsing are no problem. But throw in multimedia, and you rapidly start using up all your capacity. 3G operators have dealt with this in their unlimited-use pricing plans by disallowing these bandwidth-hogging applications in the fine print of their service agreements. But those are the very applications that consumers want. Usage-based pricing is one solution, but the market has spoken against this approach. Sprint Nextel is going to have to solve the problem of how to make money while providing sufficient bandwidth to users in a network that has relatively finite capacity. Yes, the company has some 100 MHz average capacity per market, so that 10 MHz radio channel I described could eventually become 10 radio channels. But that doesn't change the fact that there will be a certain dollars per gigabyte that it costs to deliver capacity as well as a certain number of gigabytes that users will expect to access and only so much they will be willing to pay. Bottom line: The technology choice was easy compared to what Sprint Nextel will face in the business equation. Success could allow the company to establish a leadership position in mobile broadband in the United States, but it will be a high-risk undertaking.

top^

Wide-Area Wireless--The Next Five Years
Network Computing Mobile Observer, Jul 26, 2006

Recently, I concluded the research phase of a large project to assess the future direction of all the major wide-area wireless technologies, including 3G and WiMAX. There is so much going on that it has left my head spinning. But at the same time I'm quite excited, because the future of mobile data just keeps getting brighter. As an IT manager, you won't be able to immediately take advantage of many of these enhanced technologies, but it may be helpful to know what will be available in what timeframe. It's also interesting to see how the 3G vs. WiMAX battle is shaping up.

3G today consists of the GSM/UMTS technology family and the CDMA2000 family of technologies. The version of UMTS (Universal Mobile Telecommunications System) that Cingular and other operators around the world are deploying is based on 3GPP (Third Generation Partnership Project) specifications release 5. This release includes a service called HSDPA (High Speed Downlink Packet Access). HSDPA has theoretical rates to 3.6 Mbps with today's devices, increasing to 7.2 Mbps in 2007. Real-world limitations, however, limit the peak rates you will experience to just over 1 Mbps. With multiple users active in a cell, speeds may drop to the 500 kbps range. Cingular is quoting typical rates of 400 kbps to 700 kbps. Interesting, it's only in the United States that operators quote "typical" rates. These are actually very slippery numbers, as they are based on backhaul architecture and capacity, cell site spacing, voice load and data load. Beyond HSDPA, 3GPP Release 6 specifications include a new service called HSUPA (High Speed Uplink Packet Access), which boosts peak uplink speeds to just over 1 Mbps, with theoretical peak rates of 5.76 Mbps. HSUPA will be available next year. People are calling the combination of HSDPA and HSUPA simply HSPA (High Speed Packet Access).

However, with UMTS/HSPA, it doesn't end there. A series of improvements are planned for HSPA in a combination of efforts. The first will come in the context of Release 7 functionality, and the next will come in an effort called HSPA Evolution or "HSPA+." Through a combination of radio techniques including channel equalization, mobile receive diversity and MIMO (Multiple Input Multiple Output), HSPA+ takes CDMA close to maximum theoretical efficiency. Peak rates could exceed 25 Mbps in a 5-MHz downlink/5-MHz uplink spectrum allocation and users could experience two to four times the rates they can with HSDPA today. This technology, which could be ready by 2008, could match mobile WiMAX in the same amount of spectrum.

Finally, to round out their offensive strategy, 3GPP is also busily defining an OFDM-based system called 3GPP Long Term Evolution, slated for initial deployment in the 2009 timeframe, with peak rates of 100 Mbps in a 20-MHz radio channel. The combination of HSPA+ and LTE will competitively squeeze WiMAX. Mobile WiMAX on paper exceeds HSPA performance, but it won't necessarily exceed HSPA+ by very much. Meanwhile, LTE is designed to be more efficient than Mobile WiMAX. With respect to timing, initial Mobile WiMAX networks could start appearing by the end of 2007 but are more likely in the 2007 timeframe, assuming operators embrace the technology and find spectrum to deploy it in.

If WiMAX doesn't have enough competition, it also has to watch the aggressive development moves of the CDMA2000 camp, as represented by 3GPP2 (Third Generation Partnership Project 2). The first improvement to today's EVDO (Evolution Data Optimized) Rev 0 technology will appear next year with Rev A, which boosts peak downlink throughputs from today's 2.4 Mbps to 3.1 Mbps, and boosts uplink throughputs from today's 153 kbps to 1.8 Mbps. Like HSPA, peak rates that users experience in either direction will likely be around 1 Mbps, with lower average rates. For Rev A, Sprint has indicated 450 kbps to 800 kbps average for downloads and 300 kbps to 400 kbps average for uploads. Beyond this, and available by 2007 for deployment though no operators have yet committed to this, EVDO Rev B can combine up to 15 1.25-MHz radio channels in 20 MHz of spectrum for peak rates of 73.5 Mbps. More likely, an operator would combine three channels in 5 MHz of spectrum for peak theoretical rates of 14.7 Mbps on the downlink, matching HSPA peak throughputs. 3GPP2 is also working on Rev C, which is likely to be an OFDM-based approach. One avenue here is the possible convergence of Rev C with the IEEE mobile broadband standard, IEEE 802.20. 802.20 is currently on hold because of allegations of impropriety in the standards process, but this won't necessarily affect the long-term outcome.

WiMAX proponents will have to try and prove that their technology is sufficiently better, and sufficiently less expensive, than alternatives to gain acceptance. Large companies like Intel, Motorola and Nortel are all betting heavily that they can do this. My view is that the massive investment involved in R&D and deployment--and the competitive pressures from other approaches like metro Wi-Fi--makes it likely that only two wide-area technologies can survive long term. I don't see GSM/UMTS/HSPA/LTE going away because of its dominant global market share today, meaning that five to 10 years from now, people may be trying to remember what the acronyms EVDO or WiMAX stood for.

top^

Anatomy of a Well-Designed Wireless Application
Network Computing Mobile Observer, Jul 5, 2006

Over the last month I've been evaluating some new wireless e-mail products, and it has made me think about what it means for products to be optimized for wireless. After a decade of working with wireless applications, I thought I knew all the tricks, but it turns out that there is no shortage of ongoing innovation when it comes to this medium.

I've stated in previous columns that off-the-shelf IP-based network applications generally work quite well over today's new 3G networks, such as EVDO and HSDPA, especially for lighter use. But for workers who use the applications hour by hour for work-related functions, it is very important to evaluate how reliable the application is and how it deals with inevitable wireless effects such as weak signals (which reduces throughput and increases latency), occasional disconnects and possible changes in IP address when reconnects do occur. Today's mobile middleware addresses these effects quite well. The purpose of a well-designed wireless application, however, is to address these issues directly and then some. It's instructive to look at examples.

The two e-mail products I've been working with are RIM's new 8700 and a Palm OS Treo application called ChatterEmail. Both address the smartphone form factor and both do wireless e-mail extremely well. Whereas RIM's solution is ideal for organizations with larger numbers of users because of its extensive management features, ChatterEmail is essentially an end-user product that emphasizes Internet mailbox access using either POP 3 (Post Office Protocol 3) or IMAP (Internet Mail Access Protocol). As it turns out, IMAP provides much richer functions than POP3. ChatterEmail is also in the process of making an Exchange version available.

The first obvious feature is the ability to push information, meaning that new e-mail (or other enterprise information in the case of RIM) appears automatically on the new device without any action taken by the user. RIM does this via its Blackberry Enterprise or operator-hosted server and its network operations center. ChatterEmail does this by using the IDLE command that is part of IMAP, accomplishing push without the need for an intervening server. Push is great for an always-on smartphone. For laptop applications, it might or might not make sense based on how users interact with the application.

The next feature is less obvious, namely having the application work diligently in the background to obtain information, even in the presence of network anomalies, without bothering the user with the details. RIM does this brilliantly, since it has control of the end-to-end (device to NOC) connection. From the user perspective, the e-mail just shows up, no matter how many retries may have occurred. ChatterEmail only has control of one end, resulting in a reduced ability to address networking problems. It compensates for this cleverly by doing multiple retries, even giving the user options in the settings about how aggressive it should be in attempting to connect or whether to even try in the presence of a poor signal. Before you think that maximum aggressiveness is a good thing, realize that this can result in high power consumption. ChatterEmail provides a quick command to reset network connections, which often resolves issues if e-mail is not getting through. It also provides a status in the corner of the screen of the connection steps it is going through so you can more easily diagnose issues in the case of failures.

Another area is power consumption. The RIM solution is extremely efficient in its use of the network, which helps maximize battery life. RIM can do this thanks again to control of the end-to-end connection. ChatterEmail doesn't have this luxury, as it's using an existing networking protocol that wasn't designed for power efficiency. So in its online state where e-mail appears immediately, battery consumption is relatively high, though good enough to get through a full heavy e-mail work day. But again, it compensates by having a more battery efficient polling mode, with user programmable intervals. This takes away from immediacy, but for users in meetings or in the evening, this may be just fine. A nice little feature here is that the program indicates in the
corner when it will next retrieve e-mail.

The next important area is how to deal with large amounts of information that you don't necessarily need on a mobile system, laptop or smartphone. With e-mail, this generally applies to large attachments. RIM handles this nicely by allowing you to view attachments piece by piece, so you may only download a few kilobytes to see what is in an attachment, versus megabytes of the original presentation file. Again, this requires control of both ends, with the Blackberry server intelligently dishing out portions of the attachment. ChatterEmail can't do this because it doesn't have the server component, but it cleverly provides other options for dealing with large e-mails. First, it lets you view the message without downloading the attachment; then it allows you to move the message into an IMAP folder on your server without downloading it. This way, you can keep your mailbox organized until you can deal with large messages and attachments using a full-screen computer. Both RIM and ChatterEmail provide options for deleting messages on both the device and the server.

I've only touched on the main features in this column. The bottom line is that applications tailored specifically for the wireless environment can deliver an extremely effective usage experience. Keep in mind, however, that what makes sense on a smartphone -- because of limited battery power and the device's always-on status -- may not be the optimal behavior for a laptop. If you're deploying wireless applications, I recommend familiarizing yourself with the various optimization approaches so you can evaluate applications accordingly.

top^

Using Your Cell Phone as a Modem
Network Computing Mobile Observer, Jun 14, 2006

I have been using cell phones as modems for laptops all the way back to analog cellular over a decade ago. With each phone and with each wireless technology, there has been a learning curve, sometimes a precipitously steep one. But in all cases, I have been happy with the end results. First, let's look at why you might want to do this, and then let's look at the issues with doing so given today's technologies and service plans.

Tethering options nowadays are either via Bluetooth, IR or special-purpose USB cable that connects from your laptop to your phone. In the past, cables also connected to serial ports, but USB is most common nowadays. Tethering is most useful for people who only occasionally use wide-area wireless data and who also do some amount of data on their phones. By tethering, you can use the same data plan for both the phone and the laptop. I find tethering particularly useful in a smartphone/laptop combo when traveling. I use the smartphone for viewing most of my e-mail, responding to a few urgent ones, deleting others, then leaving the rest for my tethered connection, which I may engage in at the end of the day. I'm convinced there are a lot of people running around with laptops and phones who would love to do this, and already have all the right gear, but are either not aware this is an option or, if they do, don't know where to start.

Three developments in recent years have made tethering more practical. One is that an increasing number of phones have Bluetooth capability, which I think is the superior choice for tethering because it's fast enough (though just barely) for today's 3G networks and doesn't require you to carry a cable. The second development is Bluetooth being built into many laptops. The third is a good selection of connection management applications that walk you through the initial configuration and then provide a user interface that shows the connection status and provides buttons for connecting or disconnecting from the network. Some connection managers, which are provided by phone vendors as well as cellular operators, can also manage your Wi-Fi connection. Despite some of the attractions and ready feasibility of tethering, there are a number of complications and issues that have hampered adoption.

One complication with Bluetooth is that not every phone with Bluetooth supports the necessary Bluetooth DUN (dial-up networking) or SPP (serial profile) profile. Even if the device supports Bluetooth DUN, you'll need to make sure the connection manager supports that device. The connection manager is not essential, as you can manually configure a Windows dial-up networking connection specifying a Bluetooth modem, but this borders on rocket science. If you decide to take this route, there are application notes floating around the Internet that provide step-by-step directions for different devices, so do the search rather than figuring it out yourself.

Another complication is that you generally can't use your phone if you're on a data connection. However, your phone will still ring if somebody calls, and if you take the call, your data session will suspend and then resume once your call is over. With new 3G technologies such as Universal Mobile Telecommunications System (UMTS), which Cingular is now deploying, the situation improves because UMTS allows simultaneous voice and data. This means you can be using your cell phone for a voice call even as you are using your phone as a modem for a laptop. In fact, Cingular indicated in a recent interview that the company views its 3G network as particularly well suited for tethering, though the operator also supports PC Card modems and laptops with embedded modems, depending on user preferences.

The final complication is one of pricing plans, where, until recently, operators did not have clear policies. If you're subscribing to a usage-based plan, whether for phones or laptops, things are fairly clear cut. But where things can get murky is with unlimited plans, which typically run $60 for laptops but anywhere from $10 to $40 for smartphones. Clearly, operators don't want subscribers to sign up for an unlimited smartphone plan and then consume large amounts of data with a laptop. However, operators can monitor how their data accounts are used and presumably clamp down on users abusing their accounts. And if you're already paying $30 to $40 for a wireless e-mail account with an operator, how much extra will you be willing to pay for a tethering plan? Nevertheless, the tide seems to be turning because, recently, all the major operators instituted tethering pricing plans for new devices such as the Blackberry 8700, Palm Treo 700p and Motorola Q.

Add up all the complications and, clearly, tethering is not for everybody. In a recent E-poll conducted by Network Computing to research mobile wireless broadband adoption for a forthcoming feature, only 21 percent of 593 respondents indicated tethering as their preferred approach. Nonetheless, this is still a sizeable segment of the market, and one I expect to grow.

top^

VOIP Over Wide-Area Wireless: A Tricky Proposition

Unstrung, May 25, 2006

For those of you thinking of running VOIP over wide-area broadband wireless services (e.g., EV-DO, HSDPA), here are some things to think about. In the first place, if you read the fine print of the service contract, you likely are not allowed to do so. But secondly, you may not be happy with the results. If you’ve tried it and it worked, then you were lucky, and were probably running on a relatively unloaded network. I got quite a bit of insight into this topic last week listening to an engineering presentation from Qualcomm Inc. on how VOIP will be implemented for Revision A of EVDO. The presentation was part of a PCCA meeting I chaired last week on the topic of Internet Multimedia Subsystem (IMS).

Sprint Nextel Corp. and Verizon Wireless will be deploying EVDO Rev A in the 2007 time frame. Rev A primarily features a faster uplink, but it also includes provisions such as QOS that will enable VOIP. However, that doesn’t mean that the operators will roll out VOIP right away. That will take another year or two as there are a lot of additional items required to make VOIP work.

What Qualcomm aptly showed was that to deliver high-quality, low-latency voice in IP at the equivalent (or slightly higher) spectral efficiency of current circuit-switched approaches is going to require extremely sophisticated communications protocols and voice processing. First there is packet header compression, as otherwise you’re sending 20 bytes of IP header information for every 22 bytes of VOIP payload. Robust Header compression knocks this down to 4 bytes. Then there is the elimination of PPP framing overhead, QOS implementation, de-jitter mechanisms, and items called smart blanking and time warping to recover from low-level bit loss.

Qualcomm and other vendors are actively working on these areas. Until all these capabilities are baked into the broadband wireless technology itself, any VOIP usage will be of significantly lower quality, and will consume far more bandwidth than existing voice services. Similar efforts will be required to make VOIP a reality for technologies such as UMTS and WiMax. It’s going to be the end of the decade before you see widespread VOIP over wide-area wireless networks.

top^

Convergence Update: IMS and UMA
Network Computing Mobile Observer, May 24, 2006

There is currently a lot of discussion in the industry about convergence, which is somewhat amusing because I thought we were done with convergence. The first time this buzz word came in vogue was about 10 years ago, with the rise of the Internet, especially as we were starting to inflate the big telecom and Internet bubble that drove the NASDAQ index to stratospheric heights. Well, convergence of all kinds of services, such as voice and multimedia over IP networks, never quite happened back then, despite the efforts of large companies and valiant startups alike. Now convergence is back, and this time it may just stick. I obtained considerable insights into this area from chairing a Portable Computer and Communications Association (PCCA) symposium last week on the IP Multimedia Subsystem (IMS), a convergence platform if there ever was one. The meeting was hosted by Tatara Systems and had in-depth technical content presented by Airvana, Intel, MCCI, Nokia, NMS Communications, Qualcomm and Tatara.

First, we need to clear up the term "convergence," because it actually refers to multiple things. To begin with, there is industry convergence, as companies both consolidate (e.g., wireless operators) and cross into each other's turf (e.g., Google doing metro Wi-Fi). Then thereis device convergence, with devices supporting multiple functions such as voice, video, cameras, e-mail, browsers, gaming and media players. There is also network convergence, which is the notion of people and
devices using a combination of wireless and wireless networks. Wireless network options include cellular, Wi-Fi, Bluetooth, UltraWideband, digital mobile TV broadcast, near-field communications and RFID. Finally, there is application and service convergence, where operators can centrally host services such as voice (circuit-switched or VoIP), push-to-talk, messaging, gaming, entertainment and video sharing as well as provide access to these common services regardless of network, be it cellular, Wi-Fi, DSL, coax cable or fiber to the home. IMS, in fact, is a platform for service/application convergence, which is why it is being evaluated and trialed by many large operators. However, no IMS services are deployed quite yet.

We currently have a divergence of convergence approaches. IMS is the favored long-term approach. But there is also Unlicensed Mobile Access (UMA), which is a way for GSM/UMTS voice and data services to be made available over Wi-Fi networks. In effect, UMA handles a small subset of what IMS can deliver. Consequently, it is much simpler. This is a good thing, because if you look at the IMS network reference diagram you'll be impressed by the complexity of the system, with all of its interfaces, gateways and protocols. UMA devices are already available from vendors like Nokia, and they consist of mobile phones with Wi-Fi, cellular (GSM or UMTS) and UMA client software. Although no UMA services are available quite yet, sufficient vendor commitment and operator trials mean that UMA-based services are likely to become available this year.

Why this is important is that UMA (and eventually IMS) will enable a more seamless integration of Wi-Fi and cellular. This means that you'll be able to receive cellular calls with a device operating on your Wi-Fi network. Seamless handover between the two networks will also be possible. Sure, cellular voice minutes are inexpensive in the United States, but the bigger problems this network convergence supports are coverage and, to a lesser extent, capacity. Many buildings now have good Wi-Fi coverage but poor cellular coverage, especially deep inside the building. As an IT manager looking at how to extend voice throughout your enterprise, it's not too early to include UMA on your road map of technologies to monitor. Meanwhile, the converged approach can also add to capacity, which can help in places like airports and shopping malls.

Beyond Wi-Fi/cellular and fixed/mobile convergence, many of the long-term benefits of IMS will be for operators. But the application that we're likely to see in the short term, and that should be of interest to consumers and businesses alike, is video sharing. This allows users with phones that support the feature to be on a call and, midstream, to enable video sharing. The voice call remains circuit-switched (for now), while the video travels on a simultaneous parallel data channel, probably with a slight time lag. Not only might this be useful for showing your grandma your son's antics ("isn't he cute"), it could have uses in tech support ("that yellow cable plugs into the third socket from the right"), real estate ("you'll love this house"), emergency services ("that's the license plate"), auctions ("here's the engine") and so forth. I don't have any hard dates on when video sharing will become available. However, enough of the pieces are in place that it could happen this year, and definitely by next year. These are all baby steps for convergence, but
they are real -- and they will have an impact.

top^

Palm OS Still Alive

Unstrung, May 19, 2006

After Palm Inc. ’s release of the Treo 700w using Microsoft Corp. (Nasdaq: MSFT - message board)’s Window Mobile operating system, many people were wondering what would happen to the Palm OS operating system. The fate of Palm OS was also clouded by PalmSource’s decision last year to discontinue work on Cobalt -- the high-end, multi-threaded OS that was to take Palm into the future. PalmSource indicated at that time its intention to use Linux as its high-end OS platform.

The Treo smartphone line has sold extremely well, with some 3 million Treo 600s and 650s in the market. And it’s no wonder. The Treo 650 does a lot of things well, especially from a useability point of view. So it was no surprise when Palm announced the 700p, an enhanced Palm OS Treo that will support CDMA2000 EV-DO networks. The gadget will address one of the 650’s biggest weaknesses -- namely, its limited memory. The 700p will have 128 Mbytes of memory. I’m sure Palm will sell plenty of these. A UMTS/HSDPA version seems like a logical development as well.

However, while Palm OS has its dedicated following, and is well suited for light-weight applications such as organizer functions and email, it simply does not have the horsepower of competing systems like Windows Mobile 5, which is far more sophisticated in its multi-tasking and networking capabilities. Mitigating this limitation is that most users do not do much beyond organizer tasks, email, and some browsing functions on their smartphones. So it makes complete sense for Palm OS to keep being used for the moment. However, as enterprises slowly figure out how to deploy more demanding applications for smartphones, its likely that many of these will be targeted for alternative smartphone platforms.

top^

Hotspot Redux

Unstrung, May 16, 2006

Lost in all the discussion and arguments about metro WiFi has been the deployment and usage of wireless hotspots. Remember those? Five years ago, working with Datacomm Research Co. , I published the first comprehensive market report on wireless hotspots. While we were enthusiastic about the potential of the technology, particularly in its ability to deliver broadband using relatively low-cost infrastructure, we were concerned about the ability to make hotspots a viable business unless combined with other services such as cellular.

Since then, although wireless hotspots have not taken over the mobile broadband space, the number of hotspots around the world has grown at a strong pace. According to JiWire Inc. statistics for May, there are now 114,150 hotspots in 126 countries. The number is more than double that of a year ago, which makes for impressive growth. The U.S., at 38,588, has the largest number, more than twice the next country, which is the U.K. at 16,898. Pubs, cafes, and restaurants have the lion’s share of deployment with 40 percent among them, followed by hotels at 26 percent of locations, then stores at 13 percent. And despite all predictions of free service dominating this segment, according to the stats, less than 10 percent of hotspots today are free.

Even though the market is highly fragmented, and there are no official standards for roaming, the key players in this space have done a great job of providing single-account access to multiple locations. For example, iPass Inc. (Nasdaq: IPAS - message board) boasts access to 50,000 hotspots, Boingo Wireless Inc. to 30,000, and T-Mobile USA has 6,914 of its own hotspots, with access to a total of 22,799 hotspots through roaming arrangements.

We’re still a long way from having public WiFi everywhere. It would take about a million hotspots to cover every logical location in the U.S. if you count all hotels, shopping areas, airports, restaurants, and so forth. Nevertheless, in the last five years we’ve come a long way.

top^

Mobile Broadband - Trying to Catch a Fast-Moving Target
Network Computing Mobile Observer, May 3, 2006

This April and May I've been teaching a series of one-day seminars on the evolution of Wi-Fi, 3G and WiMAX. Along with forcing me to update much of my slideware, these classes have made me assess the state of the wireless-data industry. Although wide-area wireless data revenues are increasing, they still only represent about 10 percent of operator revenues.

What's the holdup? Is what the market really wants technologically achievable? My view is that wireless networks are successful to the extent that they can act as enhancements and extensions to our existing infrastructure networks. However, one fundamental challenge is that the performance of wireline infrastructure increases very quickly. Even though wireless network performance is increasing as well, wide-area wireless networks are playing catch-up against a fast-moving target. The implications are significant, especially for the evolution of the mobile broadband market. Let's look at some numbers.

Close to 1990, when a fast dial-up modem ran at a throughput rate of 9600 bps, I examined the specifications for a forthcoming data service for digital cellular, also at about 9600 bps. My naive
view back then was that if a wireless network could offer a modem service as fast as wireline, we'd all go straight to mobile heaven. Alas, by the time the wireless service became available several years later, modem throughput rates had gone up to 28 kbps. Wireless connectivity was sluggish in comparison because of both the lower throughput rates and the higher network latencies. This trend has continued.

There was considerable excitement in the wireless industry as 1xRTT and EDGE networks were being primed for deployment. Finally, wireless would catch up! Yet by the time these technologies became available, with around 100 kbps throughput performance, most serious network users had moved to 500 kbps or higher broadband networks. Dial-up use, meanwhile, was declining as travelers started using services such as hotel-provided broadband.

Now, operators are deploying 3G networks with rates in the 500 kbps to 1 Mbps range. Have we finally caught up? It all depends on your point of comparison. Today's 3G data services are blazingly fast compared to prior wireless services. Yet compared to, say, a DSL connection, many applications operate noticeably slower.

Enhanced versions of 3G will be able to boost throughput rates by an order of magnitude too, approaching 10 Mbps of real throughput, as will mobile WiMAX. But new cable modem and DSL services are already approaching the 5 Mbps range, and with fiber to the business or the home, Internet speeds of 100 Mbps-plus become feasible. So the catch-up game will continue.

Relative to wide-area networks, Wi-Fi has done better on the speed front because it can harness more radio bandwidth per user; it also uses much simpler protocols. In fact, early IEEE 802.11n-like products are beating out 100 Mbps Ethernet in throughput tests. But Wi-Fi is a local-area technology, and despite all the buzz about metro Wi-Fi, it is ill-equipped to become the personal broadband network available anywhere.

The other wireless limitation is capacity. There simply is not enough spectrum available, even with the most spectrally efficient technologies available, to deliver to large numbers of users the gigabytes per month that many Internet users are starting to consume with applications like movie downloads. Thankfully, business users are not generally bandwidth hogs, and 3G networks should be able to satisfy their needs for quite some time.

At the end of the day, wire and fiber will win the speed and capacity battles in most populated areas. Despite this, we must realize that today's wireless networks are still immensely useful and that they will be ever more useful. The RIM BlackBerry has proved that you don't need much bandwidth to provide an extremely useful service. But this is a highly optimized solution involving gateways and a network operations center. What does it take to get beyond e-mail on a widespread basis? Do we have to do it one application at a time, as with e-mail, or are we reaching the point where the networks are fast enough and affordable enough that a whole wealth of new applications will become practical? I'm hoping for the latter.

top^

Google Eases the Third Screen Blues

Unstrung, Apr 26, 2006

Recently Google changed its search page for mobile devices so that if you click on one of the search results, Google automatically formats the subsequent page for better viewing and faster loading on a mobile screen.

This addresses the problem of large and complex Web pages that take a long time to render on a phone or PDA, and that are difficult to read despite the valiant efforts of today’s mobile browsers. I’ve used this Google capability for a while and have found the results to be quite helpful. Unfortunately, there is no way to make this rendering process perfect, because there is no algorithmic process that can extract the essential elements of a complicated Web page for a small screen.

However, because 99 percent or more of Web content was designed for large screens, the fact that Google is trying to make more of the Web accessible to mobile devices is a good thing, in my opinion. This is especially the case as network speeds are improving, phones are becoming more capable, and we’re all trying to figure out how to do more with our smartphones than just email and organizer functions.

There are some concerns about what Google is doing, such as the issue of copyright. Does Google have the right to alter the content? It’s fine with me, but then I’m not an attorney.

Another concern is what if the content has already been optimized for a small screen? (Unlikely by and large.) Actually, Google provides another search option, which is to select Mobile Web, in which case it will only report hits for pages that are formatted for mobile devices. This restricts search results to a much smaller universe, but the resulting pages should be more readable.

Overall, I think Google is onto something, and many users out there with browser-equipped phones -- especially larger ones such as BlackBerries, Treos, and Windows Mobile devices -- are likely to benefit from Google’s approach.

top^

Do-It-Yourself Wireless Email

Unstrung, Apr 18, 2006

Although there are a number of good wireless email solutions available from the likes of Research In Motion Ltd. (RIM) (Nasdaq: RIMM - message board; Toronto: RIM), Good Technology Inc. , Intellisync Corp. (Nasdaq: SYNC - message board), Seven Networks Inc. , Sybase Inc. , and now Microsoft Corp. (Nasdaq: MSFT - message board), there are also increasing options using Internet-based services. As an exercise, I decided to see if I could build my own wireless push email solution for my Palm Treo 650.

Why “push”? Couldn’t I just let the Treo poll my POP3 Internet-hosted account every 15 minutes? Well, polling does not quite provide the real-time email experience that people sometimes desire.

Here is how I did it:

First, I created a free Yahoo account. Then I arranged, via a Web page configuration option provided by my ISP for a copy of my POP3 Internet email to be forwarded to my Yahoo email account. On my Yahoo account I configured SMS notification to be sent to my Treo 650 whenever I received an email. Then on my Treo 650 I installed a free utility called “TreoHelper” in which I enabled an SMS trigger that searches for my email address in every SMS message and if it finds it, the trigger launches the Treo VersaMail email client, which then automatically downloads my email from my POP3 email account.

So that I don’t have to actually view each SMS message, I enabled an option in TreoHelper to delete the SMS after its VersaMail launch operation. The only catch now is the volume of SMS messages. At approximately 1,000 messages per month, I would need an SMS bucket plan to bring down the SMS cost to $10/month -- otherwise I’d see a $100 SMS charge.

The solution works, though I’m still waffling on using this versus the 15 minute polling approach, as 15 minutes is almost close enough to “real time” for me. Another consideration is that polling uses a little more data and battery. Regardless, the moral of the story is that there are increasing mobile options available from Internet solutions providers to accomplish what used to require dedicated wireless infrastructure.

top^

Integrated Voice/Data in 3G

Network Computing Mobile Observer, Apr 12, 2006

One of the promises of 3G cellular has been integrated voice and data. In this column I examine what this means, why and where it's important, to what extent the different cellular technologies support it, and how this capability will evolve over time. I'm bringing it up now because Cingular is currently rolling out an enhanced version of Universal Mobile Telecommunications System (UMTS) that includes High Speed Downlink Packet Access (HSDPA), which supports integrated voice and data. Meanwhile, Sprint and Verizon have recently made announcements about an upgrade to CDMA 2000 Evolved Data Optimized (EV-DO), called Revision A, that provides a key building block for integrated voice and data, a capability currently not available for EV-DO.

Integrated voice and data means that while you're on a voice call you can simultaneously engage in a data session using the same device. This allows you to do things like send pictures while on a call or do simultaneous voice and video. Not only are there personal scenarios where this might be entertaining, but this ability should prove invaluable for many job situations. Examples include insurance adjusters at the scene of damaged property communicating with headquarters, real-estate agents giving real-time interactive video tours of houses to their clients and medical technicians showing pictures/videos of people they are treating while communicating with doctors. These examples assume a handset device. In a laptop scenario, it depends on your connection. If you're using a PC Card modem or a laptop with integrated 3G capability, then you're likely in a data-only mode, where the integrated voice/data capability is not relevant. But if you're using a phone as a modem, perhaps with a Bluetooth connection, then you can make a voice call in the midst of a data session. And for those of you wanting to use a VoIP system like Skype over 3G, today's networks are simply not up to the task yet.

So which technology can do what? The most capable 3G technology in this area is UMTS/HSDPA, where handset devices support simultaneous voice and data. This is possible because the UMTS voice channels and HSDPA data channels co-exist in the same radio channel. However, in the case of EV-DO, the radio channel is dedicated to high-speed data functions while voice travels over 1XRTT (one carrier radio transmission technology) channels. The 1XRTT channels also support lower speed data services but require devices to be in either a data mode or voice mode.

The lack of integrated voice/data puts EV-DO at a competitive disadvantage relative to UMTS/HSDPA. EV-DO operators have another problem, too. With UMTS/HSDPA, the entire radio channel is available for any combination of voice/data loading. With EV-DO, operators have much less flexibility to dynamically allocate radio resources between voice vs. high-speed data. So far this has not been an issue, as EV-DO networks have relatively few users, but as data services become more popular, operators may have to add EV-DO channels; this assumes, however, available spectrum, and it also takes away from voice capacity. The answer to this dilemma is EV-DO Rev A, which operators will begin to deploy next year. EV-DO incorporates improvements that will enable VoIP functionality, including QoS support, reduced latency and improved uplink capacity. VoIP will enable integrated voice and data for EV-DO, and it will allow more flexible radio resource management, because EV-DO channels now will be able to support both voice and data functions.

However, implementing VoIP over 3G will still be a hugely complicated task. Current approaches of using dedicated CDMA channels for voice are extremely mature. Replicating the same level of voice quality and consistency in the packet domain will involve a lot of radio engineering as well as additional network enhancements such as sophisticated IP packet header compression and an
IP core for voice. An important potential element of this core will be the IP Multimedia Subsystem (IMS), a set of infrastructure elements that integrates voice, data and multimedia. IMS is based on the Session Initiation Protocol (SIP) and allows for common core services (voice, voicemail, video, presence, push-to-talk and so on) to be implemented in applications that operate independently of the access network. IMS is under consideration for GSM/UMTS, CDMA2000 and WiMAX networks as well as for wireline networks where it can facilitate wireless/fixed convergence.

I expect the path to VoIP to be a two-stage process for EV-DO carriers. The first step will be deploying the EV-DO Rev A upgrade in 2007, initially as a data-only enhancement, mostly for improved uplink speed and reduced latency. The second part will be installing the ancillary VoIP functions, which probably will not be ready for primetime until 2009. The end game is attractive, especially as it may result in additional voice capacity. VoIP is on the road map for UMTS/HSDPA as well, but the feeling of urgency in this community is much lower because it already has the benefit of a combined voice and data offering.

top^

Pay-As-You-Go Mobile Broadband

Unstrung, Apr 8, 2006

Verizon Wireless announced on March 29 a pay-as-you-go option for people buying notebooks with integrated EV-DO capability. The cost is $15 for a 24-hour period. This allows the operator to compete on a pricing basis with wireless hotspots and other temporary broadband options such as at hotels.

I applaud this pricing option and believe it will be quite popular. There are many professionals, myself included, who don’t travel enough to justify the $60 monthly unlimited plans that are currently the norm for 3G wireless-data services. Even though the new Verizon plan only takes four days to reach $60, and realizing that $60/month is only $2/day, not having the recurring monthly expense, and having the freedom to pay for the service as you need it is attractive.

With some hotels charging about $10/day for their broadband services, and others offering it for free, the service only makes sense for people wanting connectivity from a wide variety of locations. And the emphasis needs to be on “wide,” because WiFi hotspot service from companies like T-Mobile USA is only $9.99 per day, and, all other things being equal, I prefer a WiFi connection over a wide-area connection from a performance standpoint.

I am curious, however, why the service is only available for notebook computers with integrated EV-DO and not for users with EV-DO PC card modems. Perhaps it’s an interim approach for logistical reasons, or perhaps it's to encourage sales of the laptops with integrated EV-DO capability.

Incidentally, I’m a big fan of integrated 3G because of generally better radio performance, ease of setup, and ease of use. Over time, because the pay-as-you-go type of pricing plan simply makes good sense, I’m relatively confident that it will broaden to other types of devices and that it will be adopted by other operators.

top^

Mobile Middleware in the Broadband Era

Network Computing Mobile Observer, Mar 22, 2006

With mobile broadband networks like EV-DO and HSDPA becoming more widely available, many IT managers are asking about the role of mobile (or wireless) middleware. Is it still needed with the higher speeds? The answer is yes, but the reasons have changed.

For background, mobile middleware is the system that operates between a mobile device and a server. Usually it consists of software on a mobile device and a mobile server that acts as a proxy for the mobile node so it can communication with an end server such as a mail server, Web server or database server. The objective of this architecture is to optimize the communication between the mobile node and the mobile server for wireless connections, addressing such unique wireless communications challenges as slower speeds, higher and variable network latency, intermittent connections, changes in IP address because of connections being reestablished and so forth.

In its earliest forms, mobile middleware made it feasible to use really slow networks such as Mobitex and CDPD with applications that would have performed much too slowly in their native form. Today, with typical speeds of 3G networks approaching a megabit per second, speed is less of an issue. In fact, many networking applications will function quite well with no middleware assistance, especially with light use. However, the mobile environment is still fundamentally different from the high-speed Ethernet world over which many networking applications function.

I can see this with my use of a Treo 650 and POP e-mail, where about 20 percent of e-mail downloads fail because of variable delays and occasional slow throughput that confuse either the TCP or the POP3 protocols and result in timeouts. Compare this with a RIM BlackBerry architecture that is rock solid in its mail delivery, regardless of wireless network type. Why? Because it employs a mobile middleware architecture, including the BlackBerry Enterprise Server and RIM's NOC.

The same goes for the other wireless messaging gateways. With sufficient mobile design emphasis, there's no reason that direct client/server interaction cannot be improved for the mobile environment -- as Microsoft has done with Windows Mobile 5 and SP2 for Microsoft Exchange. However, this is not the case for most applications. So we are left with a situation where even Web browsing over EV-DO and HSDPA can be enhanced by a factor of two to four using Web optimization (another form of mobile middleware). With today's busy Web pages, downloading a Web page in 3 seconds instead of 10 over a 3G connection is a noticeable gain.

However, speed is just the tip of the iceberg. Another crucial element is security. Most enterprises employ VPNs for remote access, including 3G wireless use. One approach is to use a standard VPN from the likes of Checkpoint, Nortel or Cisco, which is certainly feasible. But an increasingly popular alternative is to use mobile middleware with VPN functionality (referred to by some as
mobile or wireless VPNs) from companies such as Ecutel, IBM, ipUnplugged, NetMotion Wireless and Padcom. Not only do you get the secure tunnels, but you also get transport optimization through compression. In addition, depending on the vendor, you may get important features such as session maintenance, so if you lose connections -- dropping from coverage, for example -- the mobile middleware maintains your application sessions, allowing them to quickly resume once you are back in coverage. Some mobile middleware solutions also allow users to maintain secure applications sessions even as their network changes, say from 3G to Wi-Fi.

Another increasingly important and relatively new function that mobile middleware can potentially support is policy enforcement. This is where the mobile middleware enforces what applications can access the network with what network connection type. If you only have limited bandwidth because you've roamed into a 2.5 G coverage area, it might be nice if bandwidth-hogging applications such as Windows auto-update were put on hold, letting you download your e-mail efficiently. With so much automatic networking activity occurring on today's laptops, this is a feature I can truly appreciate -- currently, I manually disable virus updates, Windows updates, spyware updates and so forth before using my wireless connection.

And what if your enterprise policy is to always use a VPN for networking, but you find yourself at a wireless hotspot where you need to use a browser outside of the tunnel to enter your hotspot login credentials? Some mobile VPNs can let you do this by temporarily allowing a browser session outside the tunnel.

The benefits of mobile middleware are quite clear, but which specific solution to use can be much less obvious. Feature sets and architectural approaches vary considerably amongst the vendors. The challenge is sorting through all the available approaches; not only are there mobile middleware vendors dedicated to this space, but the mainstream application vendors and VPN vendors are also adding mobility functions to their products.

Bottom line: If you or your users are making light use of wide-area wireless networks, you can probably get by without using mobile middleware. But if you're going to do any heavy lifting, where applications will be doing large numbers of networking transactions every day, where you may have a large number of users in the field and user productivity can only be achieved through applications that work extremely efficiently and reliably, then you seriously need to consider mobile middleware as part of your overall architecture.

top^

Wireless vs. Wireline – It’s No Contest

Mobility Loop, Mar 22, 2006

With so much emphasis on wireless networking today, you’d think that wireless was about to displace all wire. That simply won’t be the case, not because of laws of economics but due to laws of physics. When you look at speeds and capacity, you have to consider the capacity of fiber versus the capacity of radio. Fiber has a theoretical capacity in the range of 10 to 100 Tbps. That “Tbps” is terabits per second, or 1,000 Gbps. Even if you had the entire lower 10 GHz of radio spectrum available to you, and assumed a whopping 10 bps/Hz through the most advanced radio techniques available (likely breaking Shannon’s law in any real world deployment with interference), you’d still only end up with 100 Gbps. So, what we have is the entire useful radio spectrum carrying one percent of the theoretical data capacity fiber. Okay, maybe you can only do 10 Gbps over today’s fiber system, but remember, that’s just one strand. Want more capacity; add more strands. Now, take into account the tiny sliver of spectrum available to any operator, and the ratio of wireless to wireline capacity becomes even smaller.

Thanks to the telecom boom and bust, a lot of fiber was deployed, with only a small fraction of it in use today. My view is that one role of wireless is to complement the fiber in fundamentally two ways: for mobility and for filling in the gaps.

In the case of mobility, we don’t want a physical tether as we relish our ability to communicate from anywhere. So this is a perfect role for wireless, keeping in mind that the larger the cell, the more users there may be to share the spectrum. This results in typically lower average throughput rates for wide-area networks compared to local-area networks. In the case of filling in gaps, this is the other good use. In the US, copper is everywhere, but fiber only currently passes a small percentage of buildings and other target locations like broadband wireless cell sites. So while initiatives such as fiber to the curb (FTTC) are exciting, it will literally take decades to hook-up all the fiber. This is where technologies such as mesh Wi-Fi and WiMAX can help fill in the gaps. I love wireless, and see it as a technology that can transform lives and make companies much more productive, but we need to keep its capabilities in perspective.

top^

IEEE 802.11n – Dramatically Faster

Mobility Loop, Mar 17, 2006

The contention over different technical approaches for IEEE 802.11n appear to have been resolved, and the standard is moving towards completion. The IEEE 802.11n working group just gave its tentative approval to a draft 1.0 standard. Next come comments and votes from the members of the working group. If approved by 75% of members, the proposed standard will be formally submitted in July. However, based on 500 members, numerous expected comments, and the likelihood of a second version of the specification before final ratification, it will be well into 2007 before this high-speed standard is complete. And this truly will be high-speed.

As reported by Glenn Fleishman in Wi-Fi Net News on March 15, according to the CTO of Atheros, peak raw rates will be 600 Mbps with 450 Mbps achievable in the fastest configuration, with 150 Mbps realizable in a 200 Mbps base configuration. Assuming about 25 Mbps of achievable throughput for IEEE 802.11a and IEEE 802.11g, IEEE 802.11n will be six to eighteen times faster than today’s standardized Wi-Fi networks, which is a huge increase in speed.

Do we need this speed? Absolutely. Do we need it for all applications? Of course not. For Internet access where the limiting factor is the broadband connection at 1 to 5 Mbps, there will be no benefit. But for computer to computer communications locally, it will be invaluable. I regularly have to disable my Wi-Fi connection and plug in Ethernet, to do large directory synchronizations across computers. Consumer multimedia applications, especially combined with 802.11e QoS will be a big application. For instance, moving HDTV requires 20 Mbps of reliable throughput, which is touch and go with current wireless, and how many homes have just one TV screen? Also, the Multiple Input Multiple Output (MIMO) antenna technology will extend range compared to current standards, which in many cases will be as valuable as increased speed.

But be careful before you rush out and buy “n” products. Until the standard is complete and the Wi-Fi Alliance certifies products for interoperability, it could be a risky proposition buying early versions of products based on early versions of the specifications. Yes, you’ll likely get the speed, but despite what vendors may claim, there are no guarantees that pre-standard products will be upgradeable to standardized versions later on. This may be fine for smaller deployments, but potentially problematic for larger ones.

top^

TCP/IP and Wireless

Mobility Loop, Mar 7, 2006

Tuesday, March 07, 2006 Written by Peter Rysavy
In early days of wide-area wireless, using TCP/IP was a risky proposition. The first opportunity to do so was in 1994 with Cellular Digital Packet Data (CDPD), which was the first IP-based wide-area wireless network. With CDPD you could run TCP/IP applications, but the biggest problem was how slowly most of them operated. With higher speed networks like EV-DO from Sprint and Verizon, and HSDPA from Cingular, you’d think we’d be in networking heaven, and on a relative basis, we are. Certainly the throughputs are extremely good with typical quoted rates of 400 to 700 kbps. However, as networks have become faster, they still have not completely solved one problem—latency.

Latency is much better than it has ever been before, but it is still in the 200 to 300 msec range, and has high variability. In some detailed testing I did recently with one high-speed cellular network, latency across about two hundred measurements averaged 290 msec as measured by pings against a test server, but the amount of variation in delay had a standard deviation of 285 msec and the highest delay was 3900 msec. These numbers mean that the timers used in TCP’s complex algorithms for detecting network congestion can become confused, the connection can stall and effects can propagate up to the application layer where operations such as a mail download or file transfer can fail.

So yes, TCP/IP works better over wide-area wireless than ever before, but still not perfectly, and not with the same degree of reliability that you can expect with Wi-Fi or Ethernet. The bottom line is that if you are deploying a productivity application that has to be extremely reliable, you may well need to fine tune that application for wireless, or at least test it thoroughly to know that it will respond in different conditions such as weak signals or network congestion.

top^

Figuring Out Metro Wi-Fi

Network Computing Mobile Observer, Mar 1, 2006

Several weeks ago I chaired a meeting on municipal and mesh Wi-Fi developments. Since I found the meeting quite informative, I thought I would report some items I learned. If this topic seems familiar, it's because I wrote a column about it in December 2005, presenting the pros and cons of this technology. The meeting was held on February 2, 2006, in San Francisco by the Portable Computer and Communications Association and hosted by PCCA member company iPass. Leaders in the industry, including BelAir Networks, Cisco Systems, Motorola and Tropos Networks, presented on industry developments, standardization, deployments and opportunities. Combined with in-depth technical discussion, the meeting provided a fairly comprehensive snapshot on the status of the industry. Here are some interesting observations and conclusions.

Some 400 cities in the world are doing some sort of municipal deployment, so clearly this technology has gained some momentum. However, most deployments are quite recent and in relatively early stages. Mesh capability is essential in making metro-area Wi-Fi feasible. By allowing packets to hop from access point to access point, and only having a subset of access points connect back to centralized infrastructure, deployment costs are significantly reduced. The typical number of hops is between two and three. Each hop can add some 30 to 50 msec of latency. Current mesh solutions are all proprietary, but IEEE 802.11s will standardize mesh networking, allowing for interoperability across vendors. However, this standard is not expected to be finalized until 2008.

What I found especially interesting was the analysis of capacity and scalability. Typical numbers of access points range between 20 and 40 per square mile. Resulting aggregate capacity is in the 20 to 100 Mbps range, a noteworthy figure as it is about an order of magnitude higher than current cellular systems, which makes sense as there is about an order of magnitude more access points than base stations per square mile. If you want to service a large number of broadband users in a dense area, clearly Wi-Fi has a leg up on alternative wide-area technologies, assuming you can afford the required backhaul capacity.

I was also impressed by the mobility capabilities of some of these networks. The Motorola solution, designed to support both public-safety and consumer applications, can handle mobile speeds to 100 mph. Those are some really fast handoffs, and my only question is how quickly you would drive out of the coverage area at that speed. If proponents have their way, however, entire cities could be covered, and that won't be an issue. But that won't necessarily be easy. Though it may sound like a simple matter to install access points on light poles, the logistics are quite complicated: Every township has separate control over these poles, and the moment you're around freeways, the federal government has control of the lights. Network operators must involve a lot of entities to achieve widespread coverage. In addition, some poles are powered at 110 volts whereas others are powered at 440 volts. These challenges are reminiscent of the problems Metricom had rolling out its networks.

The one question that nobody was able to clearly answer at the meeting was how people would manage interference between the mesh and other networks. This remains a complicated problem because of the huge variety of radio circumstances that will arise. To minimize deployment cost, the mesh access points are located in the loosest configuration possible. If you have 25 access points per square mile, that translates to 1,000 feet between access points and relatively weak signals if you're at the maximum distance of 500 feet from the access point. This is why many of these networks call for repeaters on the sides of houses to provide indoor coverage. If you're at the weak part of the signal, your private Wi-Fi network will simply win out over the mesh signal. But if you're at the weak point of your own coverage, such as in a distant bedroom, and there is a mesh node just outside your house, that signal may prevail over yours. You may be able to switch channels if your neighbor is not already using that channel. Or the IEEE 802.11 carrier sense protocols may be sufficient for co-existence if all the appropriate nodes can hear each other. Or maybe mesh networks will stay at 2.4 GHz, where the frequency is needed for propagation, and shorter range private networks will migrate to 5 GHz. How this all shakes out remains to be seen. For public safety, government users have the option of using the licensed 4.9 GHz band, which at least isolates them from unlicensed networks.

Bottom line, I still see significant challenges for wide-area mesh networks, but their deployments have sufficient momentum that it is inevitable that there will be some level of deployment. Current lack of standardization definitely puts municipalities at risk in the event that their vendors go out of business, as has already happened to some cities that were using products from Vivato. Interoperability aside, today's vendors do have highly capable products that can support relatively broad coverage areas. However, while some people might be satisfied with the resulting coverage for fixed or nomadic applications, mobile applications are likely to need greater coverage than offered by the mesh, and so the need for much wider 3G services will not go away. As in my own use of wireless technologies for productivity purposes, I see an increasing number of applications best served by a combination of wireless technologies.

top^

3GPP Long-Term Evolution

Mobility Loop, Feb 21, 2006

Get ready for another wireless acronym – LTE. I suppose it's in the nature of wireless technology that as soon as you start widely deploying one version of a technology, you start discussing the next version. This serves to generate market excitement and keeps thousands of engineers employed working on the next big thing. Unfortunately, it can make current offerings seem inferior compared to the glowing capabilities of what is coming down the road.

The most current version of deployable 3GPP technology is Universal Mobile Telecommunications System (UMTS) High Speed Downlink Packet Access (HSDPA), as being rolled out by Cingular Wireless and other operators around the world. This is based on Release 5 of Third Generation Partnership Project (3GPP) specifications. Next year we can expect Release 6 features such as High Speed Uplink Packet Access (HSUPA), and then later with Release 7 we’ll see items like VoIP. These releases represent the short range roadmap for GSM/UMTS specifications. The longer term roadmap is what LTE is all about. LTE is the next big radio overhaul, with all the best-of-breed techniques available, including OFDM and MIMO. It will offer flexible channelization from 1.25 MHz to 20 MHz, peak theoretical downlink rates to 100 Mbps and latency down to 10 msec in the radio link. If this all sounds very similar to Mobile WiMAX, that’s not a coincidence. LTE is intended to match all of WiMAX capabilities, though WiMAX proponents would claim they are further ahead in their efforts.

Why all the attention now? Because key players in the 3G space are positioning themselves to compete in this future wireless space. For example, Nortel announced on Feb 14 at the 3GSM World Congress in Barcelona “As part of the 3rd Generation Partnership Project (3GPP) for the Long Term Evolution (LTE) of wireless network standards, Nortel has achieved approval for the introduction of high speed OFDM and MIMO packet access, in line with its technology HSOPA.” HSOPA stands for High Speed OFDM Packet Access. You can expect a lot more LTE positioning like this from other vendors. What does this mean to users in the next couple of years? Not much really, other than an indication of future network capabilities in the 2008 or later time frame. Just keep in mind that there is often a 10 to 1 ratio between the claimed peak rates and typical user rates, and that it always takes longer for new wireless technologies to become available than expected. For all practical planning purposes, stick to what is available now. In the cellular world, that’s EDGE/UMTS/HSDPA and 1xRTT/EV-DO.

top^

Sprint Nextel's Jump to the Future

Network Computing Mobile Observer, Feb 8, 2006

Many eyes are fixed on Sprint Nextel as the company contemplates how to take advantage of the large amount of spectrum it owns in the 2.5 GHz band. This is the Broadband Radio Spectrum (BRS), formerly known as Multichannel Multipoint Distribution Service (MMDS). Sprint and Nextel each had sizeable holdings, and the merged company now controls this spectrum in most first-tier markets nationally. We're talking about a sizeable amount of spectrum, close to 100 MHz in many markets, according to one of my associates who tracks these things.

Under FCC rules, the company must use it or lose it, so Sprint Nextel is closely looking at various technologies. Last year, Nextel did trials using Flarion's Flash OFDM technology, which is now owned by Qualcomm. This year, Sprint Nextel will do a WiMAX trial using Samsung's WiBro gear. Developed in Korea, WiBro is the basis of mobile WiMAX and will become one of the mobile WiMAX profiles. As such, it will be the first "version" of mobile WiMAX available. WiBro service is also being turned on in Korea this year in the 2.3 GHz band. How WiBro performs will be the first good indicator of how well mobile WiMAX will perform and whether it can outperform current 3G technologies, a hotly debated topic.

However, Sprint Nextel is not stopping with WiBro. It is also doing a trial with IP Wireless' Universal Telecommunications System (UMTS) Time Division Duplex (TDD) system. UMTS-TDD is an official 3GPP technology that is essentially High Speed Downlink Packet Access (HSDPA), but with a time division duplex radio. Just last month, Sprint Nextel invested $10 million in IP Wireless, following a $4 million investment last July. This is quite interesting given that there generally is fierce competition between 3GPP (GSM/UMTS family) and 3GPP2 (CDMA2000 family) technologies and that most operators have sided with one technology family or the other.

And to confuse things further, IEEE 802.20, the IEEE standard on mobile broadband wireless access, is now revving up again thanks to Qualcomm's purchase of Flarion and Flarion's Flash OFDM technology. Qualcomm originally saw IEEE 802.20 as a threat to EV-DO, but in my opinion it now sees IEEE 802.20 as a vehicle for competing in next-generation wireless technologies. This way, the company has its bases covered in CDMA, with EV-DO Rev A and Rev B, and in OFDM, with Flash OFDM and IEEE 802.20 standardization. I'm sure Sprint Nextel is monitoring IEEE 802.20 developments too.

Clearly, Sprint Nextel wants to keep all its options open as long as possible so the company can make the best technology decision. Whatever technology it uses, the resulting network, where deployed, will be a high-capacity high-performance network that is likely to outperform Sprint Nextel's current EV-DO network. The company has indicated it will be deciding on the technology this year, making Sprint Nextel the first major wireless operator in the United States to commit to a next-generation wireless technology.

The technology choices are fascinating, but it will be especially interesting to see how Sprint Nextel chooses to market and price the service once it is broadly available in the 2008 time frame. Will the company use it to off-load data from its EV-DO network? You can make a good argument for that, as there is only so much capacity in the Sprint 3G network -- roughly 1.5 Mbps per cell site per EV-DO radio channel, which is not that much. Or will Sprint Nextel use it to compete with DSL and cable? This is a trickier proposition because these users demand far more capacity, anywhere from 300 megabytes per month to 1 gigabyte per month on average based on my calculations, though I will confess that I have found it challenging to get good data. (Contributions on this subject are gladly accepted.) Or will Sprint Nextel emphasize multimedia applications? Opportunity also exists for local bypass of telephone companies, which must be a tempting target, including local phone service and T1 circuits for businesses.

Although I understand the technology options, I'm a lot less clear on what the successful business models are in the wireless broadband area. I do truly believe there is a mass market for mobile broadband, whether delivered to laptops, smartphones or mobile multimedia devices. But when you factor in the huge expense of deploying a nationwide network, the large amounts of data that multimedia applications can consume, the finite amount of capacity in wireless connections compared to wireline alternatives, and the limited amount of money that users are willing to pay, is there a business model that actually works? A lot of people, including myself, are going to be watching Sprint Nextel closely over the next several years to find out.

top^

Mobile Content – Getting Better Slowly

Mobility Loop, Feb 3, 2006

When I find myself having to wait in places like bank lines, post office lines, or alone at a restaurant, the first thing I do is whip out my Treo smartphone and surf the Web. This usually keeps me occupied for about fifteen minutes, before I’ve read much of the available content. Then I’m out of luck, because the choices are so limited. Whereas the World Wide Web on a regular computer for all practical purposes contains an infinite amount of content, the amount of entertaining or interesting content formatted for small screens is much more limited.

I usually use Yahoo (http://wap.oa.yahoo.com/) because the site does a fairly good job formatting for small screens, is well organized, and has a reasonable amount of news. In fact, M:Metrics, a company that does mobile-oriented market research, just announced in a press release on January 30 that Yahoo is the leader in this space, with 12.8 million US mobile subscribers accessing the site in December 2005. AOL is second place with about 9 million, MSN and Google following with about 7 million, and ESPN the sports leader with about 5 million.

The operators themselves have portals. I sometimes use Cingular’s MEdiaNet site (powered by InfoSpace), but find that it has some rough edges, such as displaying text in only the left half of the screen. Overall, we've come a long way with mobile content, and tens of millions of subscribers is a great start, but in the big picture of things, only ten percent of mobile phone subscribers are accessing mobile content, so we still have a long way to go in both subscribers and in the depth and breadth of the content. Clearly, the level of effort by content providers is commensurate with the amount of access, hence the relatively minor offerings so far. I’m personally looking forward to the day that I can spend a couple of hours reading a whole issue of New York Times. That way I’ll be covered no matter how long a wait in line I have.

top^

The Resurrection of IEEE 802.20

Mobility Loop, Jan 25, 2006

Last week, IEEE 802.20 met in Hawaii, with companies making significant technical contributions. All indications are that IEEE 802.20, nicknamed Mobile-Fi, is being brought back from the dead. What’s changed? Everything. IEEE 802.20 was initially a vehicle that Flarion Technologies used to try and standardize its Flash OFDM technology. But it never made much standards progress due to lack of cooperation from other participants who had other agendas, including Qualcomm which at the time viewed Flash OFDM as a competitive threat to CDMA2000 EV-DO.

However, now Qualcomm owns Flarion, and is pursuing opportunities to advance its technology portfolio across all future wireless fronts. This includes EV-DO (which it invented), UMTS/WCDMA (where it is a leading chip supplier), and OFDM-based technologies (where it now has a reasonable intellectual property-rights position with the Flarion acquisition). Qualcomm now is pushing IEEE 802.20. If it actually achieves standardization and gets to market, IEEE 802.20 will compete directly with mobile WiMAX (IEEE 802.16e).

Despite the immense size of the global wireless opportunity, I don’t see the market as large enough for three versions of 3G (EV-DO, UMTS and the Chinese TD-SCDMA), the evolved versions of 3G (e.g., 3GPP Long Term Evolution), WiMAX, mesh Wi-Fi and now IEEE 802.20. Something has to give.

From a standards point of view, WiMAX is much further along than IEEE 802.20. However, Flash OFDM is a proven, tested technology, though its not necessarily exactly what is being proposed for IEEE 802.20. (I haven’t had a chance to figure out the differences but will try to find out for a future blog.)

The interesting question to me is what effect IEEE 802.20’s revival will have on the mobile broadband market. For example, it could prove disruptive to WiMAX if operators decide they need to evaluate one more wireless technology option.

top^

The Smartphone Conundurm

Network Computing Mobile Observer, Jan 18, 2006

Palm's official unveiling of its Windows Mobile Treo smartphone, the Treo 700w, at the beginning of this month made me contemplate the current status of smartphones and their likely evolution. Overall, I'm enthusiastic about this platform, but I have to confess that from a productivity point of view, it's easy to invest just as much time working with these devices as any time you will save from using them. That's the conundrum.

First, let's look at the some of the recent smartphone developments. The first Windows Mobile 5 devices have now appeared. These include the Treo 700w, available from Verizon Wireless and the PPC-6700 from Sprint, both with stylus support and small keyboards. Meanwhile, Cingular started selling its Cingular 2125, which is based on Windows Mobile 5 but more compact -- it uses a regular telephone keypad instead of a stylus and keyboard. There are more Windows Mobile 5 PDAs without
phone functionality than there are with phones for the simple reason that adding wireless to these devices is complicated, and wireless operators have rigorous acceptance procedures. Windows Mobile 5 will have a big impact as it addresses some of the serious shortcomings of the prior version, including volatile memory and the need for two-handed operation with a stylus. Anyone who has spent much time with a RIM Blackberry of Palm Treo understands the importance of one-handed operation. The Treo 700w even shows the battery level, a reading that required multiple steps with the Pocket PC 2003 platform. Meanwhile, connectivity speeds are increasing with RIM Blackberrys, which now support EDGE and EV-DO. RIM also recently released its attractive 8700 line. And PalmSource has decided to use Linux for future platform development. Finally, the Symbian platform gets its best shot for visibility in the United States with the Nokia 9300 being sold through Cingular.

Smartphones are intriguing, as they represent efforts to concentrate the greatest amount of computing and networking power possible in a small form factor that is comfortable to carry. The result is tremendous capability, with a good range of productivity or entertainment applications, such as e-mail, Web browsing, Web applications, file access, portable documents, database access, music, video and hundreds more (just search the online catalogs). For most workers, smartphones can't replace a laptop, but the multitude of features present makes it increasingly possible for many workers to use their smartphone for short trips or other periods away from the office.

However, this level of capability also means complex computer systems and all the associated maintenance and reliability issues. Although I'm reasonably happy using an EDGE-capable Treo 650, I have now updated the system software twice to newer versions as they became available. I was glad each time, because the updates addressed various problems. Although Palm documented the update steps clearly, these were still complex operations that likely defeated many users. I also spent a considerable number of hours getting Bluetooth to work for various usage models, such as using the phone as a modem, for HotSync and with my wireless headset. And still, on a regular basis, the phone will lose awareness of the headset, and I have to pair the two devices again. I've searched various support forums and have found other users with the same problem, so I don't think it's just me. As for stability, it's not bad, but not perfect. My device runs for about a week before I have to reset it. These are just examples of my usage. If you want to see the scope of support issues for these devices in general, just read the thousands of entries in the various support forums for the different platforms.

When I add up my support time, I have to question how much further ahead I am from a productivity point of view. From an IT perspective, these devices must be managed; their software must be kept up-to-date for potentially large numbers of devices, preferably through some centralized system; and remote-access security architectures must be updated to support smartphones, all adding up to considerable time investments. And if you don't make these infrastructure investments, you'll pay even more in compromised security and supporting devices on a case-by-case basis.

Still, I see the ascendancy of the smartphone platform as inevitable. One reason, after functionality, is that prices are dropping, with good smartphones now available in the $200 to $300 range, putting them in reach of a lot of people. Gartner in a July 2005 report predicted global sales of 200 million smartphones by 2008, over 20% of all mobile phones sold. However, unless smartphone vendors want to see an IT backlash against these devices for wasting too much support time, they should make sure their platforms and included applications are as stable as possible. It took quite a few releases of Microsoft Windows before it finally became reasonably stable with Windows 2000 and Windows XP. After this long painful experience, IT and user tolerance for system instability has gone down, so let's make sure we don't have to repeat this process with smartphones.

top^

Mobile Phone Music

Mobility Loop, Jan 12, 2006

There’s no question that people want to carry fewer devices. Witness the struggling PDA market thanks to smartphones that now include PDA functionality. I used to carry a phone and iPAQ Pocket PC, and could even do nifty things like download e-mail to my iPAQ via a Bluetooth connection to my phone that acted as a packet-data modem. But I’m much happier with a single device, currently a Treo 650. Similarly, I don’t carry a separate music player but play MP3 songs on my Treo 650, either using headphones or through a Radio Shack tape deck adapter in my car. Though the Treo 650 is less than ideal for music, the concept is powerful, and further evidence that our “phones” are becoming mobile electronic life enhancers.

A lot of companies are now pursuing music on cell phones, though I would say we are still far from musical mobile nirvana. Motorola has their “ROKR” phone, available from Cingular, that is iTunes compatible, but only stores 100 songs. Sprint has a music download service, but songs cost $2.50 each. Most recently, Verizon last week announced at the Consumer Electronics Show their Vcast Music service where users can download songs to a handset at $1.99 each (a little pricey), or 99 cents to a computer. The controversial twist here is that the specific mobile phones supporting this service use Windows Media player. While the phones can play MP3 format songs, once users invoke the service, the MP3 capability is disabled on the phones. However, if a user already has MP3 songs, Windows Media player will transcode the songs to Windows WMA format, so they can be played. We’ll see how users respond to this, as even iPods can play MP3 format songs. I believe that any friction like this, similar to cellular’s failed attempts in the past with walled-garden content, discourages widespread adoption.

Still, one million available songs and Verizon’s marketing push will likely have a big impact. I believe that one way or another, a lot of people are going to sign up for music on their cellphones. Another big enabler that will begin rolling out this year will be Bluetooth stereo headphones that also support phone voice functions.

top^

The Perils of the Bleeding Edge

Mobility Loop, Jan 5, 2006

In wireless networking, there is a fine balance between proprietary systems that provide enhanced capabilities and the safety of standardized solutions supported by multiple vendors. Nowhere is this more evident than in the expanding field of public Wi-Fi systems where vendors are using multiple approaches to extend the reach of Wi-Fi over wide areas, in essence trying to morph a local area technology into a wide area technology. One of the pioneers in this area was Vivato Inc., which used a beam forming radio technology to dramatically improve range. Unfortunately, at the end of last month Vivato went out of business, leaving a number of municipal Wi-Fi networks built with Vivato’s technology asking questions about how to maintain their networks.

The current most popular approach for municipal networks is not Vivato’s approach, but to use mesh technology, where access points don’t all have to be connected to a backbone. Access points forward packets from other access points towards nodes on the backbone, thus significantly reducing overall deployment costs. Though leading vendors such as Cisco have gotten into the game, helping legitimize the mesh concept, all current solutions are vendor specific, putting municipalities that use these products at risk should their vendors go out of business. Fortunately, mesh Wi-Fi standardization is coming, through IEEE 802.11s. However, it is too early to tell what impact this standard will have, and to what extent risks of vendors going out of business will be mitigated. It is one thing to have a Wi-Fi network in a building become obsolete. It is another matter entirely when the Wi-Fi network covers an entire city.

For those of you wanting to learn more about municipal and mesh Wi-Fi technical developments, I recommend a meeting I’m chairing for the Portable Computer and Communications Association on Feb 2, 2006 in San Francisco that will investigate this topic in detail.

top^

HSDPA Speed: Lab versus Real World

Mobility Loop, Jan 2, 2006

Recently I complained about the inflated speed claims being made about a future version of EV-DO, (EV-DO Revision B). So it’s only fair to comment on the misleading speed claims coming out about HSDPA. While I am enthusiastic about the capabilities of both the major 3G technologies, whether EV-DO or UMTS/HSDPA, I think it’s important to understand the basis of the claims, and what can actually be expected in the real world. A case in point is Nortel’s announcement on December 22 about achieving a 3.6 Mbps HSDPA throughput rate using a card from European-based Option in a test conducted at Nortel’s research campus in Chateaufort, France. High Speed Downlink Packet Access (HSDPA) is the version of UMTS that Cingular is currently deploying.

Here’s the deal. The 3.6 Mbps number is real, but you as a user will never see it. It is based on using the highest order modulation available (16 Quadrature Amplitude Modulation), and five high-speed codes allocated to the same device. Even then, it is at the link layer, meaning that it does not take into account IP and TCP packet overhead, though this is likely less than 10%. It is effectively the capacity of most of the cell site directed at one user under optimal radio conditions. But even if you were the only user in that cell sector with an excellent signal, while the radio could deliver this throughput, the bottleneck in the real world is the backhaul connection to the base station, which is typically based on T1 circuits that are limited to 1.5 Mbps of throughput. I can guess that Nortel has a fiber connection to their test cell site in Chateaufort, but that’s not the real world, though to be fair, some cell sites are on fiber. But most are not.

The reason you’ll see actual throughputs in the 400 to 700 kbps, as quoted by Cingular, is two fold. One, as just mentioned, is the real world constraint of backhaul capabilities, which by the way is no different for Wi-Fi hotspots. The other is that multiple users will be active in the cell, sharing the radio capacity. So beware of the inflated speed claims. While they may represent the real capabilities of individual devices, and the theoretical capabilities of the technology, they will not necessarily be what you will experience.

top^

Municipal and Mesh Wi-Fi

Network Computing Mobile Observer, Dec 21, 2005

Despite significant challenges, an increasing number of cities are planning to deploy Wi-Fi networks across large areas. High-profile cities currently include San Francisco, Philadelphia and Houston, but as described in a new report titled "2006 Muni Broadband Outlook" from Light Reading Insider, other cities include Anaheim, Calif.; Atlanta; Austin, Texas; Boston; Buffalo, N.Y.; Chicago; Denver; Kansas City, Mo.; Lenexa, Kan.; Los Angeles; Miami Beach, Fla.; Milwaukee; Minneapolis; Oklahoma City; Phoenix; Pittsburgh; Portland, Ore.; Sacramento, Calif.; St. Paul, Minn.; and Seattle.

There is a considerable amount of debate over whether these deployments are going to become the favored form of broadband wireless access or are doomed to failure. It's a good question, as a case can be made for both sides. That's one reason why the Portable Computer and Communications Association, where I chair the standards and architecture committee, will be holding its next meeting on the topic of municipal and mesh Wi-Fi. We will have technically oriented presentations from the leaders in the field, including BelAir Networks, Cisco, Motorola and Tropos Networks, and good representation from the computer, networking and wireless industries. The meeting is on Feb. 2, 2006, in San Francisco, hosted by iPass, and is open to anybody who wishes to attend, though there is a meeting fee for non-member companies.

Allow me to muse on the cases for and against. The biggest case "for" is one of capacity, as I've already written about in various reports. It's a matter of trivial mathematics that sharing spectrum over a small geographic area versus over a far larger cell site results in a much larger capacity per user. This is because that same spectrum can be re-used over and over. The result is that Wi-Fi can deliver some two orders of magnitude greater bits per second per square meter than cellular data networks. Simply consider a one-kilometer radius cell site versus a 100-meter radius Wi-Fi access point, and that's a hundred-to-one difference in coverage area.

Because of the simpler CSMA protocols and smaller networks, Wi-Fi also has inherently lower latency, less than 10 msec, whereas 3G networks are struggling to get down to 100 msec. As a result, Wi-Fi provides a better broadband experience and scales better for dense user configurations. Another pro is the prevalence of Wi-Fi-equipped computers, whether notebooks or handhelds. The architecture of mesh makes a lot of sense, too, namely that only a subset of access points need a connection to a wireline infrastructure. With standardization coming through IEEE 802.11s, we'll eventually have interoperable mesh infrastructure. And with VoIP becoming a reality for Wi-Fi, even voice service becomes a possibility.

On the "against" side of the equation, the primary issue is one of spectrum management. For reasonable coverage, mesh deployments favor the 2.4 GHz band using IEEE 802.11b or IEEE 802.11g. However, there are only three non-overlapping channels in this band (1, 6, 11). One common approach to continuous coverage of an area is to use all three channels, effectively using up the whole 2.4-GHz band. Any privately deployed network, or other public network, will have to co-exist in the same radio channels. The CSMA protocols allow for this to some extent, but only if all the nodes involved can hear each other--not always the case for nodes inside a building. How bad is the problem? It's hard to say.

In discussing this with my colleague Dave Molta, his observation is, "If the mesh router is in close proximity (operating at maximum-allowed transmit power), you're likely to experience significant co-channel interference. However, if the signal level has attenuated significantly, as is often the case (in fact, that's one of the big problems with mesh deployments that try to get acceptable in-building penetration from externally mounted routers), then your local signal will overpower the signal from the mesh router. You may still experience some level of co-channel interference, but local performance will probably be adequate." How big of an issue interference is remains to be seen, but we'll find out soon enough.

Then there are the logistics of deploying a huge number of access points over a wide geographic area as well as managing and maintaining them. There will also inevitably be holes in coverage, especially indoors. Whether these systems can be made reliable enough for critical applications such as public safety, other than in select areas, is doubtful. As for VoIP, this typically requires an additional 10 dB SNR compared to data, thereby requiring even more access points.

Bottom line: I see Wi-Fi becoming increasingly prevalent in public areas. Given the unlicensed spectrum and multiple entities using the spectrum as well as the inherent problem of providing continuous coverage with low-powered transmitters, however, I don't see these networks spreading everywhere. Rather, I see them as providing patchworks of coverage in select areas like downtown and in some communities. Nor will they replace 3G data capabilities, which will always be able to provide much better continuous coverage over much larger areas, though at somewhat slower throughput rates. I continue to see public Wi-Fi as highly complementary with 3G. We are quickly moving to a world of multiple wireless access networks. Knowing which to use for what applications will be the tricky part.

top^

Ultra Wideband Standardization

Mobility Loop, Dec 15, 2005

I've always been a big fan of Bluetooth, though I must admit I have wasted more hours getting Bluetooth connections to work between my various peripherals than with any other networking technology. Currently I use Bluetooth for syncing my Treo 650 to my PC, for using my Treo as a modem for my laptop and for connecting my Bluetooth headset. It's quite evident that personal area wireless makes a great deal of sense, and so I anticipate good things for Ultra Wideband (UWB) over the second half of this decade with its blistering rates up to 1 Gbps. That way I'll be able to sync my PDA in one second. Actually, better uses will be to connect all the high speed, multimedia peripherals working their way into our digital homes; for example no more messy cable going to the high definition TV.

Unfortunately UWB has been deadlocked in IEEE 802.15.3 standardization efforts, with competing Multiband OFDM and direct sequence approaches unable to reconcile with each other. The OFDM approach is championed through the WiMedia Alliance and backed by a lot of companies, including HP, Intel, Kodak, Microsoft, Nokia, Philips, Samsung, Sony and TI among others.

It was notable when the WiMedia Alliance announced on December 8, 2005 that it had been able to get its Physical and Media Access Layers approved as UWB standards through the European based industry association Ecma International. Ecma is a non-profit organization that has issued standards in near-field communications, DVD interchange standards, scripting languages, communications technology and consumer electronics. To download the new standards, go to this site and this one.

Ecma standardization does not carry the "weight" of IEEE standardization, but given that IEEE standardization for UWB appears to be going nowhere, Ecma standardization definitely gives MB-OFDM UWB legitimacy. You can expect products to start rolling out in 2006. For higher-level interfaces, MB-OFDM UWB will use USB, IEEE 1394 and even Bluetooth, facilitating product implementation using existing driver and application architectures. And if some of the Bluetooth type of issues such as managing pairing relationships are properly addressed (a big if), UWB should have a bright future.

top^

Mobile Broadcast Technology Developments

Mobility Loop, Dec 7, 2005

Verizon Wireless last week announced it is going to be the first major carrier to support QUALCOMM’s MediaFLO technology. This announcement is indicative of some important developments in the wide-area wireless space, namely the deployment of mobile broadcast technologies to complement 3G and evolved 3G networks. Though services such as V Cast from Verizon and equivalent services from other operators provide distribution of video and music already, there is an inherent disconnect between the amount of bandwidth that multimedia can consume (particularly with video content) and the limited capacity of cellular-data networks. It is far more efficient to have a single stream of data received by multiple subscribers than separate data streams to each individual user. Hence all the activity. The relative capacities of these networks is an item I studied for a recent report titled “Hard Numbers and Experts’ Insights on Migration to Evolved 3G and 4G Wireless.”

QUALCOMM’s MediaFLO technology is based on OFDM, with FLO standing for Forward Link Only. QUALCOMM is actually building out its own national network that cellular operators can contract with, using former UHF channel 55 in the 700 MHz band. Devices will need separate radios to access both the 3G and FLO networks. The network is expected to deliver up to 20 live streaming video channels and 10 audio channels, and also some short on-demand programming.

There are a variety of other broadcast and multicast technologies in the works. There is Digital Video Broadcasting Terrestrial (DVB-T), for which a number of Mobile TV networks are being deployed globally, and a mobile-optimized version called Digital Video Broadcasting-Handheld (DVB-H). There is also EV-DO Platinum Multicast which allows multicast slots within EV-DO channels, and Multimedia Broadcast Multicast Service (MBMS), which will allow multicast and broadcast channels within UMTS/HSDPA.

While it’s not entirely clear how many subscribers will be willing to pay how much for these emerging multimedia services, it is clear there will be no shortage of wireless technologies to support them.

top^

Linksys WRT54G -- Update

Network Computing Mobile Observer, Dec 7, 2005

Since our story last week about Cisco/Linksys changing the WRT54G Wireless-G Broadband Router from a Linux-based system to a WxWorks-based system, Linksys contacted us to say it will maintain the Linux-based system under a new model number of WRT54GL. The company's statement was, "We are honored by the overwhelming loyalty of Linux users to our products, and this dedicated SKU is our way of saying thank you for the support this community has given to us and our products."

We are pleased that the Linux-based version (current equivalent to WRT54G v4) will remain available for those who desire it. However, we're still not happy about the apparent quality and performance issues of the currently shipping WRT54G. Also, it's a little unusual to create a new SKU for an existing product and to use the old SKU for, effectively, a new version of the product. However, it does mean that Linksys/Cisco can direct its huge volumes of this product's sales, estimated at hundreds of thousands per month, to the version that costs the company less to manufacture. Hopefully, Linksys/Cisco can quickly address the problems with this version so that these hundreds of thousands of users don't have the unfortunate experience I had.

top^

Linksys WRT54G -- Not What It Used to Be

Network Computing Mobile Observer, Nov 30, 2005

The Cisco Linksys WRT54G is one of the most popular SOHO 802.11 b/g wireless routers available. However, because of a recent redesign, this product is nowhere near as reliable as it used to be. Worse yet, it's difficult to tell whether the version you are getting is the old version or the new version.

I stumbled across this a couple of weeks ago when I decided to update my home small-office router, a Linksys BESFR41 v1. Though functional, I had to reset the product once or twice a week, and having had good success with a WRT54G (router, four-port Ethernet switch, 802.11b/g access point) at another location, I decided to get one for my office. As part of my due diligence, I contacted my local ISP, which has well-informed technical support, and asked them about the WRT54G. They assured me it was "rock solid." That was good enough for me.

After obtaining the product, a WRT54G v5, I installed it and instantly found my first problem: extremely slow Web admin screen updates that took five to 10 seconds to update. I then looked at the firmware version: 1.00.0. That was scary. Then I downloaded the latest version of the firmware, release 1.00.2, dated Nov. 1, 2005. This fixed the slow admin page updates, and I completed the installation. It was then that the real problems began. The router ran fine for a while, then I lost wireless connectivity but still had wireline connectivity. I checked all the router wireless settings and found that the MAC address access list I had enabled had disappeared. A one-time glitch, I thought, so I reentered the information. Things continued to work for another several hours. Then I lost Internet connectivity again. I reset the router, which restored operation. However, from that point forward, I had to reset the router at least a couple of times a day. Another problem occurred when I tried to change wireless security settings. I was unable to do so without first restoring the router to its factory settings--not exactly convenient and anything but rock solid.

I then started researching these problems on the Internet and found no shortage of complaints about the product, including stability and throughpu